SA: Various fixes to sa.js to make it work in JDK8
rednaxelafx at gmail.com
Wed Jul 31 07:11:57 PDT 2013
I've just learned that there's a bug number that's exactly what I was
trying to fix with this patch:
Bug 8011888: http://bugs.sun.com/view_bug.do?bug_id=8011888
If there's already a fix to this bug then I'm fine with it. Just like to
see the fix in mainline tip so that CLHSDB would work out of the box again.
On Mon, Jul 29, 2013 at 3:53 PM, Krystal Mok <rednaxelafx at gmail.com> wrote:
> Hi all,
> Could I have a couple of reviews for this patch, please?
> project caused a few issues that stopped sa.js from working. This is a
> patch that tries to fix the issues that I've run into.
> I'll walkthrough the patch below: (line numbers refer to sajs.patch, not
> Line 8-12:
> property name in the "dot" syntax is not a problem anymore. So I'm
> un-commenting these lines.
> Line 20-42, 51-76:
> When implementing a JavaAdapter for SA ScriptObject, the __get__ function
> calls the __has__ function:
> which is working at the wrong level: __has__ is a special hook function,
> and cannot be called via "this" this way. It'll trigger the __call__ hook
> to find the "__has__" member, but the JavaAdapter here does not override
> __call__, and then the lookup will fail.
> Rather than going through the trouble of implementing the __call__ hook
> just for this purpose, I move the __has__ function up, and made __get__
> call __has__ directly instead. Now it won't trigger the __call__ hook for
> the lookup, the things will work fine again.
> Line 45-48:
> Removed trailing whitespace.
> Line 84-85:
> This code used to work in Rhino, but apparently Nashorn doesn't
> JS-to-Java interop, so this conversion has to be done manually.
> Line 93-104:
> I'm a little confused in this part. Nashorn exposes a "print" function
> that's defined in jdk/nashorn/api/scripting/resources/engine.js, and that
> it doesn't expose a corresponding "println" function. I'm not sure if this
> code really worked when using Rhino...anyway, the "println" function is
> missing, so "writeln = println" doesn't do anything useful. I'm adding a
> shim here just in case either of "print" or "println" functions are
> missing. This fixes an error when calling "println" in line 87 of this
> Line 113:
> This is the same change as purposed by Yunda back in April.  A missing
> fix from the NPG changes.
> Line 121-122, 129-130:
> When specifying a method overload in Nashorn, the argument syntax it takes
> for inner classes uses "." as the separator between the enclosing class
> name and the inner class name, instead of "$" as in the "binary name".
> Line 138-144, 152-160:
> Nashorn has stricter default behaviors than Rhino when overriding Java
> methods. Where as Rhino defaults to a "do nothing" implementation, Nashorn
> defaults to throwing UnsupportedOperationException for unimplemented
> Line 168-172:
> Before the fix, this code only replace the first occurrence of the
> specified special characters, which happens to be not enough for newer
> HotSpot types in C++ templates.
> That's all for this patch.
> BTW, there is another patches pending review, too: JDK-8011979 
More information about the hotspot-dev